gdcmanon(1) GDCM gdcmanon(1)
NAME
gdcmanon - Tool to anonymize a DICOM file.
SYNOPSIS
gdcmanon [options] file-in file-out
gdcmanon [options] dir-in dir-out
DESCRIPTION
The gdcmanon tool is an implementation of PS 3.15 / E.1 / Basic Application Level Confidentiality Profile (Implementation of E.1.1 De-
identify & E.1.2 Re-identify)
This tool is split into two very different operating mode:
o An implementation of PS 3.15, see -e and -d flags
o A dumb mode, see --dumb
Dumb mode and PS 3.15 do not work well together, you should really only use one type of anonymization. In case of doubt, avoid using
--dumb.
In order to use the PS 3.15 implementation (-d & -e flag), you'll need a certificate to do de-identification operations, and the associated
private key to do the re-identification operation. If you are only doing a one-shot anonymization and do not need to properly re-identify
the DICOM file, you can safely discard the private key and only keep the certificate. See OpenSSL section below for an example on how to
generate the private key/certificate pair.
gdcmanon will exit early if OpenSSL was not configured/build properly into the library (see GDCM_USE_SYSTEM_OPENSSL in cmake).
PARAMETERS
file-in DICOM input filename
file-out DICOM output filename
or
file-in DICOM input directory
file-out DICOM output directory
options
You need to specify at least one operating mode, from the following list (and only one):
Required parameters
-e --de-identify De-identify DICOM (default)
-d --re-identify Re-identify DICOM
--dumb Dumb mode anonymizer
Warning when operating in dumb mode, you need to also specify an operation to do, such as 'remove' or 'empty' a tag, see below the dumb
mode options.
options
-i --input DICOM filename / directory
-o --output DICOM filename / directory
-r --recursive recursively process (sub-)directories.
--continue Do not stop when file found is not DICOM.
--root-uid Root UID.
--resources-path Resources path.
-k --key Path to RSA Private Key.
-c --certificate Path to Certificate.
encryption options
--des DES.
--des3 Triple DES.
--aes128 AES 128.
--aes192 AES 192.
--aes256 AES 256.
dumb mode options
--empty %d,%d DICOM tag(s) to empty
--remove %d,%d DICOM tag(s) to remove
--replace %d,%d,%s DICOM tag(s) to replace
general options
-h --help
print this help text and exit
-v --version
print version information and exit
-V --verbose
verbose mode (warning+error).
-W --warning
warning mode, print warning information
-E --error
error mode, print error information
-D --debug
debug mode, print debug information
environment variable
GDCM_ROOT_UID Root UID
GDCM_RESOURCES_PATH path pointing to resources files (Part3.xml, ...)
Typical usage
De-identification (anonymization, encrypt)
The only thing required for this operation is a certificate file (in PEM format).
$ gdcmanon --certificate certificate.pem -e original.dcm original_anonymized.dcm
Re-identification (de-anonymization,decrypt)
The only thing required for this operation is a private key (in PEM format). It is required that the private key used for the re-
identification process, was the actual private key used to generate the certificate file (certificate.pem) used during the de-
identification step.
$ gdcmanon --key privatekey.pem -d original_anonymized.dcm original_copy.dcm
You can then check that original.dcm and original_copy.dcm are identical.
Multiple files caveat
It is very important to understand the following section, when anonymizing more than one single file. When anonymizing multiple DICOM
files, you are required to use the directory input. You cannot call multiple time the gdcmanon command line tool. Indeed the tool stores in
memory during the process only a hash table of conversion so that each time a particular value is found it get always replaced by the same
de-identified value (think: consistent Series Instance UID).
Dumb mode
This functionality is not described in the DICOM standard. Users are advised that improper use of that mode is not recommended, meaning
that important tag can be empty/remove/replace resulting in illegal/invalid DICOM file. Only use when you know what you are doing. If you
delete a Type 1 attribute, chance is that your DICOM file will be not accepted in most DICOM third party viewer. Unfortunately this is
often this mode that is implemented in popular DICOM Viewer, always prefer what the DICOM standard describes, and avoid the dumb mode.
The following example shows how to use dumb mode and achieve 5 operations at the same time:
o Empty the tag (0010,0010) Patient's Name,
o Empty the tag (0010,0020) Patient ID,
o Remove the tag (0010,0040) Patient's Sex
o Remove the tag (0010,1010) Patient's Age
o Replace the tag (0010,1030) Patient's Weight with the value '10'
You are required to check which DICOM attribute is Type 1 and Type 1C, before trying to 'Empty' or 'Remove' a particular DICOM attribute.
For the same reason, you are required to check what are valid value in a replace operation.
$ gdcmanon --dumb --empty 10,10 --empty 10,20 --remove 10,40 --remove 10,1010 --replace 10,1030,10 012345.002.050.dcm out.dcm
Multiple operation of --dumb mode can take place, just reuse the output of the previous operation. Always use gdcmdump on the input and
output file to check what was actually achieved. You can use a diff program to check only what changed (see diff(1) for example).
Irreversible Anonymization
In some very case, one would want to anonymize using the PS 3.15 mode so as to take benefit of the automatic conversion of all content that
could contain Patient related information.
In the end all Patient related information has been removed and has been secretely stored in the 0400,0500 DICOM attribute. However to make
sure that no-one ever try to break that security using brute-force algorithm, one want want to remove completely this DICOM attribute. This
will make the DICOM:
o Completely free of any Patient related information (as per PS 3.15 specification)
o Remove any mean of people to brute force attack the file to find out the identity of the Patient
In this case one could simply do, as a first step execute the reversible anonymizer:
$ gdcmanon -c certificate.pem input.dcm anonymized_reversible.dcm
and now completely remove the DICOM attribute containing the secretly encrypted Patient related information:
$ gdcmanon --dumb --remove 400,500 --remove 12,62 --remove 12,63 anonymized_reversible.dcm anonymized_irreversible.dcm
OpenSSL
On most system you can have access to OpenSSL to generate the Private Key/Certificate pair.
Generating a Private Key
Command line to generate a rsa key (512bit)
$ openssl genrsa -out CA_key.pem
Command line to generate a rsa key (2048bit)
$ openssl genrsa -out CA_key.pem 2048
Command line to generate a rsa key (2048bit) + passphrase
$ openssl genrsa -des3 -out CA_key.pem 2048
Generating a Certificate
From your previously generated Private Key, you can now generate a certificate in PEM (DER format is currently not supported).
$ openssl req -new -key CA_key.pem -x509 -days 365 -out CA_cert.cer
DICOM Standard:
Page to the DICOM Standard:
http://dicom.nema.org/
The DICOM Standard at the time of releasing gdcmanon is:
ftp://medical.nema.org/medical/dicom/2008/
Direct link to PS 3.15-2008:
ftp://medical.nema.org/medical/dicom/2008/08_15pu.pdf
Warnings
Certains attributes may still contains Protected Health Information (PHI) after an anonymization step. This is typically the case for
Patient's Address (0010,1040). The reason is that this particular attribute is not supposed to be in the composite IODs in the first place.
DICOM Supp 142 includes it (however gdcmanon does not implement it).
SEE ALSO
gdcmconv(1), gdcmdump(1), gdcminfo(1), openssl(1)
COPYRIGHT
Copyright (c) 2006-2011 Mathieu Malaterre
Version 2.2.0 Tue Feb 5 2013 gdcmanon(1)